Could a Second SVB Crisis Emerge? High-Risk AI Trading Standardizing Investor Strategies Puts Regulators Worldwide on Alert

AI-driven trading rapidly proliferates, alongside mounting fraud schemes
Algorithmic trading patterns risk becoming homogenized, raising warnings over investor “herding behavior”
Fears grow over a replay of the SVB collapse, prompting regulatory crackdowns worldwide

“AI trading” — the use of artificial intelligence to design securities trading algorithms and automate transactions — is spreading at a rapid pace. Retail brokerages are increasingly rolling out related systems, driving up individual investors’ reliance on AI. Experts, however, warn that if AI trading standardizes investor behavior, the resulting market shock could be severe. In particular, analysts caution that in periods of heightened volatility, concentrated waves of sell orders and fund withdrawals targeting specific firms could trigger a “second Silicon Valley Bank (SVB) crisis.”

AI Technology Reshapes the Securities Market

On May 26, Nikkei Asia reported that major Asian retail brokerages are positioning AI capabilities as their next-generation weapon for retaining retail clients. In March, Hong Kong-based fintech company Futu Holdings launched “Skills,” an AI assistant built on DeepSeek’s large language model (LLM). Users can integrate Skills with AI models such as OpenClaude, Claude, and Cursor to issue trading strategy instructions through natural language prompts. Meanwhile, Tiger Brokers, a securities brokerage platform operating in Hong Kong and Singapore, reported that cumulative global interactions with its proprietary “TigerAI” surged 500% within a year of its March launch. TigerAI functions as an AI-powered investment assistant capable of reading, summarizing, and analyzing market data on behalf of investors within the brokerage platform.

The problem is that as AI-based investing becomes mainstream, side effects are becoming increasingly pronounced. One of the most visible issues is fraudulent activity exploiting AI technology to lure investors. In one example, a South Korean firm advertised that it could generate profits through AI-powered arbitrage trading between cryptocurrency exchanges and distribute daily returns to investors. Posing as a legitimate company holding cryptocurrency-related licenses, the firm instructed unspecified individuals to deposit certain digital assets. Investigations later revealed that the operators had no physical office or identifiable executives and were conducting business entirely online. They also encouraged multi-level investment recruitment schemes by promising additional rewards tied to the number of participants and investment size. The structure was ultimately identified as a classic Ponzi scheme funded by new investor deposits.

In January, prison sentences for the entire management team of Popcorn Soft were finalized after they deceived investors by promising outsized returns through an AI trading program. Between March 2022 and July 2023, the executives held investment seminars nationwide, claiming that investments in domestic and overseas futures markets through their proprietary AI trading program would guarantee 15% returns on principal. The scheme amassed nearly $88 million in investor funds across more than 30,000 transactions. Investigators found that the company had never generated stable trading profits and had instead operated using a Ponzi-style structure. The so-called AI trading program itself turned out to be little more than an automated execution tool operating on preconfigured human inputs.

The Dangers of Algorithm-Driven Herding

Another core risk associated with AI trading is the potential homogenization of investor trading patterns. If market participants move simultaneously based on signals generated by similarly structured AI algorithms, stock prices could repeatedly surge or collapse irrespective of corporate earnings or underlying fundamentals. Such “herding behavior” among investors could inflict massive disruption on financial markets. A representative example illustrating this danger was the collapse of Silicon Valley Bank in the United States in 2023.

Founded in 1983, SVB grew rapidly through a business structure deeply intertwined with the startup ecosystem. When startups deposited fundraising proceeds, the bank recycled those funds into loans for other startups, positioning itself as a key financial intermediary within the venture industry. SVB counted approximately 44% of U.S. tech startups among its clients, including high-profile firms such as Spotify and Roblox. As of June 2021, deposits in Silicon Valley totaled $62.2 billion, roughly double the $31.5 billion held by second-ranked Wells Fargo.

The situation reversed during the COVID-19 pandemic. Amid ultra-low interest rates and a venture investment boom, SVB experienced a sharp increase in deposits and expanded its asset allocation into long-duration securities such as U.S. Treasuries and mortgage-backed securities (MBS). However, when the Federal Reserve aggressively raised interest rates thereafter, prices of existing long-term bonds plunged, rapidly swelling unrealized losses on SVB’s holdings. At the same time, the venture investment market cooled, worsening startups’ cash conditions. Unable to secure fresh funding, startups began withdrawing deposits en masse, forcing SVB to liquidate loss-making bonds to raise cash.

Ultimately, SVB incurred a $1.8 billion loss after selling $21 billion worth of securities in March 2023. Markets interpreted the move as a liquidity crisis, and panic spread rapidly after prominent venture capital firms advised portfolio companies to “pull deposits immediately.” Startup clients linked through online and mobile banking platforms rushed to withdraw funds simultaneously, generating $42 billion in withdrawal requests in a single day. It was a “digital bank run” unfolding at a far faster pace than traditional banking crises. Unable to secure sufficient liquidity to respond, SVB was shut down by U.S. regulators. To prevent contagion across the broader financial system, the U.S. government and Federal Reserve announced full protection for SVB deposits along with the emergency Bank Term Funding Program (BTFP).

Governments Tighten Their Response to AI Trading

Experts warn that the risks posed by AI trading could become even more pronounced if market volatility intensifies further. Under normal market conditions, prices are formed through competition among diverse investment strategies. In an AI trading environment, however, heightened volatility could drive concentrated trading in the same direction. If AI algorithms simultaneously detect the same risk signals, massive sell orders or withdrawal requests could converge at once, amplifying market shocks almost instantaneously. In extreme cases, this could generate a disruption comparable to a “second SVB crisis.”

Professor Keith Lee, who teaches AI and financial markets at the Gordon School of Business under the Swiss Institute of Artificial Intelligence, explained that such herding behavior may appear rational from the standpoint of individual investors seeking to sell declining assets. However, if all investors move identically and the behavior spreads across the broader market, asset prices could collapse to zero across multiple classes, potentially triggering a systemic financial crisis. He referred to this phenomenon as the “Hirshleifer effect.” Lee added that in financial markets, the Hirshleifer effect describes situations in which individually rational decisions generate significant social costs at the system-wide level. Since manifestations of this phenomenon were observed during the 2008 global financial crisis — including in program trading and derivatives margin calls — it has remained one of the key concerns of financial regulators worldwide under the broader framework of “systemic risk” over the past two decades.

This concern explains why regulators worldwide are escalating oversight of AI trading. Hong Kong’s Securities and Futures Commission (SFC) recently designated the use of AI for investment recommendations and research as an official “High-Risk Practice,” requiring licensed financial institutions to implement immediate risk mitigation measures. Singapore’s Monetary Authority (MAS) likewise released consultation papers on “AI Risk Management Guidelines,” stipulating that financial institutions must independently assess AI-related risks before deployment and ensure systems remain within controllable boundaries. China has already imposed a comprehensive ban on the use of AI tools for work purposes by securities firm employees, citing data privacy and execution risks. One Shanghai-based brokerage was fined approximately $280,000 for failing to disclose the limitations of its AI-powered investment recommendation algorithm.

The UK Parliament also warned in a recent report that broader adoption of AI within finance could simultaneously heighten consumer harm and financial instability. The report identified cybersecurity threats, dependence on a small number of Big Tech and cloud providers, and algorithmic clustering effects as key risks. It further recommended that the Bank of England and the Financial Conduct Authority (FCA) introduce stress testing frameworks designed specifically for the expansion of AI-driven investing.

Meanwhile, the U.S. Securities and Exchange Commission (SEC), through its “2026 Examination Priorities,” announced plans to intensively scrutinize the impact of automated investment tools and AI trading on retail investors, particularly individual and elderly investors. The initiative reflects concerns that as AI technology becomes deeply embedded across investment decision-making processes, unforeseen market distortions must be proactively contained. South Korea, for its part, is verifying the stability of automated investment algorithms and investor protection mechanisms through the “Robo-Advisor Testbed” system led by financial IT solutions provider Koscom. Under the framework, firms must maintain a minimum standard of algorithmic risk management and internal control systems before launching related services.